[enhancement] Extended FixedPoint library for quantized ML models development

### What's hard to do? (limit 100 words)

I am using DXLS to write a backend for another tool that compiles ML models directly to HLS code. FixedPoint types are a hard requirement to deploy quantized models. The current library is missing functionality (e.g., fmas, dot products), while also having a resource overhead due to the reason in [Issue1848](https://github.com/google/xls/issues/1848).

### Current best alternative workaround (limit 100 words)

Currently, I am working on a copied version of the original fixed_point.x standard file where I add all the required functionality. I am considering rewriting all core functions without the `struct` to avoid the overhead inquired by the current type system.
e.g.,
```C
fn dot_prod
// unsigned exp DOT PROD
(x: FixedPoint[VEC_SZ],
y: FixedPoint[VEC_SZ])
-> FixedPoint {

for (i, acc): (u32, FixedPoint) in u32:0..VEC_SZ {
fmadd_already_widened(x[i], y[i], acc)
}(fixed_point::make_fixed_point(sN[NB_DOT_PROD]:0))
}
```

### Your view of the "best case XLS enhancement" (limit 100 words)

TI2 to unblock [Issue1848](https://github.com/google/xls/issues/1848) and contribute by providing the identified missing functionality required to deploy quantized models.